Anthropic has moved the 1 million-token context window for Claude Opus 4.6 and Claude Sonnet 4.6 from beta to general availability, opening the feature to all users while keeping pricing unchanged. The company said the same rate applies from the first token through the millionth token, with no tiered pricing.
Under the published pricing, Opus 4.6 remains at $5 per million input tokens and $25 per million output tokens. Sonnet 4.6 stays at $3 for input and $15 for output. That pricing matters for developers handling very large codebases, lengthy research materials, and long contract files in a single prompt instead of splitting them into smaller chunks.
MRCR v2 results show a wide gap
Anthropic also released benchmark results tied to long-context performance. Using the MRCR v2 framework to measure retrieval and reasoning across very long inputs, Claude Opus 4.6 scored 78.3%, compared with 36.6% for GPT-5.4 and 25.9% for Gemini 3.1 Pro.
The gap is large. Anthropic’s release also pointed to a distinction between raw context capacity and actual comprehension quality: while the Gemini line is known for supporting up to 2 million tokens, the MRCR v2 results suggest that a bigger window does not automatically translate into stronger performance on long-context understanding tasks.
Claude Code gets the feature across developer workflows
On the product side, the 1 million-token context window is now fully enabled in Claude Code. Anthropic said this should ease context compression issues in large repository analysis and log searches, reducing the need for manual input splitting and lowering the chance that important information gets cut off.
The company also said a single request can process up to 600 images or a 600-page PDF. The stated use cases include reviewing large sets of technical documents, checking long audit reports, and analyzing the full source code of an open-source project in one pass.